The financial system is mostly made of promises backed by promises, with arbitrarily deep levels of recursion. This is not obvious, in part because most of us get our first exposure to finance with that increasingly rare object, the bearer instrument: if you're holding a dollar bill, you have a dollar, whether you earned it, borrowed it, found it, or stole it. But physical currency is a small share of the total amount of money in the US financial system: the US has about $2.3tr of physical currency outstanding, but over $21 trillion of "money" defined more broadly as checking and savings accounts, CDs, repurchase agreements, etc.

In the years leading up to the financial crisis, one feature of financial discourse was debating the risks of derivatives exposure. There were over-the-counter (OTC) derivatives referencing $600 trillion of assets outstanding in 2007, or about ten times GDP.¹ The worry at the time was that this implied a staggering amount of leverage in the system, and there simply wasn't enough collateral to back it up.

But there was a natural rejoinder. Suppose a big bank gets worried that interest rates will rise, and so it makes a bet that pays off based on the gap between current interest rates and rates a year from now, referencing a notional value of $1bn. So if rates go up 1%, the bank makes $10m; if they drop 1%, it loses $10m; if they're flat, nothing happens² . The first thing to note about this is that the bank is hedging, i.e. the notional value of the trade offsets risks they're already running. The second don't-worry-so-much argument revolves around that bank's counterparty and how its behavior changes the numbers: let's say some time passes, and the counterparty wants to exit the trade. So they call up the bank to see if they have any interest in exiting the trade, but the bank says no. Fortunately the counterparty isn’t out of luck! They can make the same directional bet as the bank, but with someone else—another counterparty.

So our original counterparty made one bet, with $1bn in notional value, that interest rates will go down (because it's taking the opposite side of the bank, which was betting that rates would go up), and it made another bet, also with a $1bn notional value, that rates will go up. So its total economic exposure to interest rates from those two bets is exactly $0: every $1 it makes on one trade is precisely offset by $1 they lose on other trades. But they got that way by having $2bn in exposure.

This is very much a feature of the OTC derivatives market, i.e. trading derivatives peer-to-peer instead of through a centralized exchange and clearinghouse. If you make the same bet through stock options, you don't need to track down the seller to negotiate an exit; you have standardized contracts that you can sell on an exchange.³ OTC derivatives markets will naturally have higher notional value relative to their economic risk compared to other markets, because the simplest way to close a trade is often to make another trade.

But that doesn't mean the trillions of dollars of exposure are worth ignoring—what if one of the counterparties defaults, and is no longer able to fulfill its end of the deal on a pair of offsetting trades? Now, both of their counterparties suddenly have unhedged risk.

This is not so much a description of what happened in 2008 as it is a description of what was being avoided by the policy response to 2008 (and what many market participants were suddenly and realistically worried about). Suppose you run a fixed-income portfolio for a bank subject to the constraint that, if you're buying some esoteric asset like AAA-rated tranches of subprime-backed collateralized debt obligations, you must hold lots of capital for those trades—but you need a lot less if you simultaneously buy insurance from a highly-rated counterparty like AIG. AIG ran the numbers and found that default probabilities for those esoteric assets were astronomically low, so they wrote lots of insurance. The yield on CDO-plus-insurance was higher than the yield on equivalently rated corporate debt or treasury bonds, but the risk profile looked similar: it was a safe asset, with insurance provided by a safe insurer.⁴ But if that insurer runs into trouble, suddenly your fixed-income portfolio requires more capital, and the portfolio manager needs to sell, which probably triggers another round of the same.

When big banks were getting bailed out, and the market was being backstopped by extra liquidity from central banks, the policy goal wasn't to reward banks for taking risks—it was to avoid the chaotic situation where nobody knew what their portfolio was because they didn't know which of their counterparties were sound. It's possible to operate in such a situation, but not with the kind of leverage investors were using in 2008. So you can view one aim of post-crisis policy as trying to extend an inevitable process of reducing counterparty risk: instead of having a major node in the network evaporate overnight, they'd keep it alive long enough to reduce its centrality in a graceful way.

Most investors do not spend most of their time thinking about counterparty risk, for good reason. If you're an investor, it's a risk you face, but one that is very tightly regulated: if you own 100 shares of Ford, what you really own is your broker's promise to deliver 100 shares or their cash value if you sell. But your broker backs that promise by actually buying the stock. Similarly, your bank account isn't technically money that you have, but instead it’s money that your bank owes you—and there's plenty of regulation to ensure that they'll have the money if you ask for it.

Modern capitalism functions much more smoothly when promises of assets are fungible with the real thing, but much easier to transfer—which makes them easier to borrow against, increasing the share of the economy's assets that consist of promises to deliver other assets. And we're implicitly participating in this all the time: rent or a mortgage is a promise, and it's backed by the (less certain) promise of a regular paycheck. The promises-stack allows us to capitalize the future and shift consumption around optimally; it would be annoying to have a world where everyone lives at home with their parents until their thirties or forties, when they've finally saved up enough to buy a house for cash. The promises model is better, but it's more complex, and an economy made mostly of promises will run into a crisis when those promises start to break. Worrying about that is legitimate, and is a concern for both investors and regulators—but the other thing to worry about is that restricting the promise-based economy slows growth, and also implies a world where nobody can be trusted to do what they said they would.

Read More in The Diff

We’ve covered counterparty risk and its effects in many contexts in The Diff: 

• Modern Financial History Begins in 1998 examines the consequences of the Fed’s response to the collapse of Long Term Capital Management (and of many other companies making the same trade). When it was written, that piece described how markets have worked since—but now it describes how markets worked from 1998 to 2022.

• Lending and the Prisoner’s Dilemma ($) is a piece from early in the pandemic, looking at how banks’ default rates are partly a function of other banks’ lending.

• In Clean Trades and Dirty Hedges, we look at the difficulty of making one trade that perfectly offsets another and still leaves room for profits, in markets and more broadly.

• The Banks: What’s Next? A more recent look at the same dynamic where the banking system’s stability depends in part on banks continuing to lend through difficult times.

• Understanding Archegos ($): the collapse of highly levered tech investor Archegos Capital is a case where people rounded counterparty risk down to zero and then suddenly realized that it dominated returns.

1. Derivatives always look scarier when they're quoted in terms of their notional value, i.e. the value of whatever they're referencing, because some of the biggest derivatives markets are for the slowest-moving assets. You don't need a lot of leverage if you're trading penny stocks, and your broker won't give you much. But if you're betting on risk-free short-term interest rates, you need a lot of leverage to have any meaningful impact on your portfolio.

2. It’s worth mentioning that the notional value here is typically very large compared to the amount of money actually at risk, and only loosely correlates with it. But it’s also worth noting that a trade referencing a low notional value doesn’t mean that losses are bounded by that value: sell call options on $1m in notional value of a stock that subsequently goes up 10x and your initial notional exposure is a wild underestimate of the total risk.

3. This was not always the case. At least one minor market crash in the 19th century was precipitated when a railroad director defaulted on options contracts he'd personally written.

4. As it turned out, AIG did not realize any losses on this insurance: they went under because the market prices of the assets they insured dropped, and they had to put up more collateral for their trades, which they were unable to do.

Help us learn more about you!

We're building a media kit and need some anonymous aggregated subscriber data. If you work in finance and have time for a 3-question survey, that would be very helpful:

Join the discussion!